Field
Embodiments of the present invention generally relate to network interconnect technologies, such as Ethernet networks. In particular, embodiments of the present invention relate to improving network scalability.
Description of the Related Art
Over the past several years, the computing and storage server industries have been migrating towards a network-based computing and storage model to take advantage of lower cost, high-performance commodity processors and lower cost, high-density storage media. This server industry trend has created a need for a highly scalable interconnect technology to enable the various computing and storage resources to be efficiently and effectively coupled. One type of interconnect that has been considered for this purpose is an Ethernet network.
An Ethernet network is a loop-free switching path, reverse path learning network. By “loop-free”, it is meant that there is only one path between any pair of hosts (e.g. nodes) in the network. Because of this loop-free property, it is possible for the switches in an Ethernet network to forward packets by broadcast flooding, and to populate their forwarding tables through reverse path learning.
Specifically, when an Ethernet switch encounters a packet with a destination host address that it does not have in its forwarding table, the switch broadcasts that packet on all outgoing links, except for the link on which the packet was received. All subsequent switches that receive the packet that do not have the destination host address in their forwarding tables do the same thing. Eventually, the packet will be delivered to the destination host. Because there is only one path to the destination host, it is assured that broadcasting the packet in this way will not create an infinite loop of broadcasts.
In addition to broadcasting the packet, a switch also determines, from the packet, the address of the source host that sent the packet. It also notes the link on which the packet was received. This address and link association is stored in the forwarding table of the switch. In the future, if the switch receives a packet having the address of the source host as the destination address, it will know, based upon the address and link association in the forwarding table, which link to switch the packet to. It will not need to broadcast the packet. In this way, an Ethernet switch learns the reverse path of a packet. Because of this reverse path learning capability, it is not necessary to pre-configure the forwarding tables of Ethernet switches. The switches can build these forwarding tables on the fly. This self learning capability of Ethernet switches is a key “plug and play” attribute of an Ethernet network, and is one of the reasons why Ethernet is so widely deployed.
As noted above, a switch stores in its forwarding table an address and link association for each source host address that it encounters. Because a forwarding table can store only so many entries, the number of host addresses that it can accommodate is limited. This in turn limits the number of hosts that can be accommodated by the overall network (since each host has its own unique address). Thus, the scalability of Ethernet networks is currently limited by the size of the forwarding tables in the switches.
This limited scalability has been made even worse by recent developments, which have caused the number of hosts (and hence, the number of host addresses) to proliferate. These developments include the sharing of a large capacity link by multiple hosts, and the advent of virtual machines. In the link sharing context, a single high capacity link (e.g., a 10 G bit link) is shared by a relatively large number of hosts, each of which is allocated a small portion of the link's capacity (e.g., ten hosts, each allocated 1 G bit). Because of the sharing, the link has not just one host but multiple hosts coupled to it. This in turn means that the link will have multiple host addresses associated therewith. As a result, the switches in the network will need to store multiple host addresses in their forwarding tables to accommodate the packets going to and coming from that link. If a fair number of the links in the network are shared links, thereby meaning that a fair number of the links have multiple host addresses associated therewith, the storage capacity of the forwarding tables of the switches in the network will be quickly consumed. This can lead to the network being unable to accommodate all of the hosts that are coupled to it.
Virtual machines create the same problem. The concept behind virtual machines is that a single physical machine can be transformed into multiple virtual or logical machines. Each virtual machine acts as a separate machine and has its own unique address. A single physical machine can be transformed into any desired number of virtual machines. Thus, if a physical machine is transformed into a large number of virtual machines, and if that physical machine is coupled to a link (thereby meaning that all of the virtual machines are coupled to that link), then the link will have a large number of virtual hosts and, hence, a large number of host addresses associated therewith. This in turn means that the switches in the network will need to store multiple host addresses in their forwarding tables to accommodate the packets going to and coming from that link. As was the case with shared links, this can cause the storage capacity of the forwarding tables of the switches in the network to be quickly and completely consumed, which can lead to the network being unable to accommodate all of the hosts that are coupled to it. Overall, the proliferation of host addresses can significantly reduce the scalability of an Ethernet network.